Test cascade validation warns when cascade functions are missing

Test cascade validation with class-based evaluator

Test cascade validation handles syntax errors gracefully

Test cascade validation with evaluator that has cascade functions

Test checkpoint behavior at various boundaries

Test checkpoints occur at correct intervals with real evolution

Test error handling for non-existent checkpoint

Test checkpoint save and resume with real LLM

Test that initial program is not re-added when resuming from checkpoint

Test that saving/loading preserves island feature maps correctly

Test that checkpoints include best program information

Test that child inherits parent's island when no target_island is given

Test that child is placed in target_island when explicitly specified

CRITICAL TEST: This tests the fix for issue #391. When we want to add a child to island 1 (empty), but the parent came from

Test that compilation failures are captured as artifacts

Test that concurrent modifications to current_island cause issues This simulates what happens in _submit_iteration when multiple work

Test that concurrent submissions don't cause race conditions

Test that max_snapshot_artifacts is accessible from database config

Test that loaded configs have required sections

Test that evolution trace config is properly integrated

Test that the config file loads without errors

Test that config files load without errors

Test that the config file loads

Test that to_dict includes early stopping settings

Test configuration validation works

Test config with both temperature and top_p set (OpenAI compatible)

Test loading config with null temperature uses default

Test loading config with null top_p

Test config with only temperature set (typical for Anthropic)

Test config with only top_p set

max_line_len parameter should control line truncation

max_lines parameter should control block truncation

Test using tracer as context manager

Test that controller initializes correctly

Test that the controller properly extracts and stores file extension

Test starting and stopping the controller

Test that convergence_threshold defaults to 0.001

Test setting custom convergence_threshold

Test that custom complexity and diversity from evaluator override built-in calculations

Test setting custom initial_changes_description

Test setting custom early_stopping_metric

Test setting custom early_stopping_patience

Test setting custom system_message_changes_description

Test loading custom max_snapshot_artifacts from config dict

Test DatabaseConfig default for max_snapshot_artifacts

Test that database current_island is properly restored after sampling

Test creating database snapshot for workers

Test that database state is unchanged after sampling from island

Test that the database can store and retrieve programs

Test default feature dimensions are complexity and diversity

Test that max_snapshot_artifacts defaults to 100

Test _determine_program_type with feature_dimensions parameter

Test that _direct_evaluate still supports dict returns

Test that _direct_evaluate supports EvaluationResult returns

Test that disabled changes description works without diff_based_evolution

Test LRU eviction in diversity cache

Test diversity score caching functionality

Test diversity feature calculation in feature coordinates

Test diversity reference set updates

Test that programs with identical content are not added multiple times

Test exploitation mode with empty archive

Test handling of empty feature_stats

Test that empty inspirations list doesn't break anything

Test that empty island falls back to sample()

Test that empty islands are initialized with copies, not shared references

Test getting top programs from empty island

Test full flow: eval failure -> artifact -> prompt -> next gen

Test that old evaluators still work unchanged

Test EvaluationResult with artifacts

Test that evaluator.py works

Test the evaluator with a simple program

Test that the evaluator file exists

Test that evaluator exists

Test that evaluator exists

Create a test evaluator file

Test that the evaluator has an evaluate function

Test that the evaluator has required functions

Test that Evaluator can load the function_minimization evaluator

Test that evaluator_models list is also updated after rebuild

Test that function_minimization evaluator has cascade functions

Test that evaluators return proper EvaluationResult objects

Test that the evaluator can evaluate the initial program

Test that evaluator creates temp files with correct suffix

Test that the evaluator uses the correct file suffix

Test that all evaluators can be imported without errors

Test EvolutionResult dataclass

Test that EvolutionResult has the expected structure

Test that evolution works with small local model (may not be perfect but should not crash)

Test evolve_algorithm with simple class

Test evolve_code with string input

Test evolve_code with real optillm server - code string optimization

Test evolve_function with simple test case

Test that evolve_function generates an evaluator that works in a subprocess. This is a regression test for the bug where callable evaluators

Test evolve_function with real optillm server - simple optimization task

Test that examples have the minimum required files

Test that explicit target_island overrides parent island inheritance

Test that explicit target_island works even with fallback parent

Test that exploitation mode samples from archive

Test that sample_from_island respects exploration/exploitation/random ratios